Setting aggregated gradient scaling factor to maximum gradient norm performs best for task balancing

{
  "doc": "lin-2023-dual-balancing.md",
  "context": "Empirical finding on choice of αk in gradient normalization strategy",
  "category": "ai",
  "norm_label": "Setting aggregated gradient scaling factor to maximum gradient norm performs best for task balancing",
  "graphify_id": "claim_max_norm_choice",
  "source_file": "lin-2023-dual-balancing.md",
  "imported_from": "/Users/antonborzov/Documents/Research.nosync/papers/extract_typed_out/lin-2023-dual-balancing/graph.json",
  "extracted_type": "claim",
  "source_location": "§3.2",
  "graphify_file_type": "claim"
}